Errors while compiling Hercules under AIX 5.3

[aix_tech]

Hi everyone,

I am attempting to compile Hercules under AIX 5.3 and came
across the following errors, during the "make" phase.
 
hstructs.h:1133: error: field 'mtget' has incomplete type
hstructs.h:1147: error: field 'stape_getstat_mtget' has incomplete type make: 1254-004 The error code from the last command is 1.

Can someone give comment on this?

Thank you

Victor

[kerravon86]

--- In hercules-390@..., "aix_tech" <aix_tech@...> wrote:
>
> I am attempting to compile Hercules under AIX 5.3 and came
> across the following errors, during the "make" phase.
>  
> hstructs.h:1133: error: field 'mtget' has incomplete type
> hstructs.h:1147: error: field 'stape_getstat_mtget' has incomplete type make: 1254-004 The error code from the last command is 1.
>
> Can someone give comment on this?

Did you switch on OPTION_SCSI_TAPE?  Try switching
it off. My guess is that it's not supported on
your system.

BFN.  Paul.

[fish-8]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

aix_tech wrote:

> Hi everyone,
>
> I am attempting to compile Hercules under AIX 5.3 and came
> across the following errors, during the "make" phase.
>
> hstructs.h:1133: error: field 'mtget' has incomplete type
> hstructs.h:1147: error: field 'stape_getstat_mtget' has
> incomplete type make: 1254-004 The error code from the last
> command is 1.
>
> Can someone give comment on this?

Hi Victor!

Check your config.h header and tell me what value "HAVE_SYS_MTIO_H" has.

If it's not defined then your system doesn't support SCSI tape (or if it
does then you don't have all the pieces installed).

If that's the case then you'll have to manually #undef OPTION_SCSI_TAPE in
"hostopts.h".

We probably shouldn't be checking "__GNUC__" in hostopts.h. That's likely
our bug. Just because someone is using the GNU C compiler doesn't mean the
system it's being used on (or the system the program is being built FOR) is
necessarily a "Linux" system! You can blame me for that I guess.

Does anyone know of a better preprocessor symbol to use instead of
__GNUC__??

- -OR- ....

Perhaps maybe we should just add the following near the end of hostopts.h?

  #if !defined( HAVE_SYS_MTIO_H )
    #undef OPTION_SCSI_TAPE
  #endif

Would that maybe be a better fix??

- --
"Fish" (David B. Trout) - fish@...
Fight Spam! Join CAUCE! <http://www.cauce.org/>
7 reasons why HTML email is a bad thing
http://www.georgedillon.com/web/html_email_is_evil.shtml
PGP key fingerprints:
DH/DSS: 9F9B BAB0 BA7F C458 1A89 FE26 48F5 D7F4 C4EE 3E2A
RSA: 6B37 7110 7201 9917 9B0D 99E3 55DB 5D58 FADE 4A52




-----BEGIN PGP SIGNATURE-----
Version: PGP Desktop 9.8.0 (Build 2158)
Charset: us-ascii

wj8DBQFKyL2XSPXX9MTuPioRAif7AJ9qkh6MjOJq6PcqxCQSTp78qY+l9wCdG1UT
h/KAuaZdojQgQp4UzheZgDQ=
=Kku4
-----END PGP SIGNATURE-----

[kerravon86]

--- In hercules-390@..., "Fish" <fish@...> wrote:
>
> system it's being used on (or the system the program is
> being built FOR) is necessarily a "Linux" system! You can
> blame me for that I guess.
>
> Does anyone know of a better preprocessor symbol to use instead of
> __GNUC__??

No, but I can tell you that when I went looking in
the gcc executable on my Linux system for something
to use for conditional compilation, I went with:

defined(__gnu_linux__)

> - -OR- ....
>
> Perhaps maybe we should just add the following near the end of hostopts.h?
>
>   #if !defined( HAVE_SYS_MTIO_H )
>     #undef OPTION_SCSI_TAPE
>   #endif
>
> Would that maybe be a better fix??

Perhaps both?

BFN.  Paul.

[fish-8]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

[...]
> No, but I can tell you that when I went looking in
> the gcc executable on my Linux system for something
> to use for conditional compilation, I went with:
>
> defined(__gnu_linux__)

Done! Thanks.


- --
"Fish" (David B. Trout) - fish@...
Fight Spam! Join CAUCE! <http://www.cauce.org/>
7 reasons why HTML email is a bad thing
http://www.georgedillon.com/web/html_email_is_evil.shtml
PGP key fingerprints:
DH/DSS: 9F9B BAB0 BA7F C458 1A89 FE26 48F5 D7F4 C4EE 3E2A
RSA: 6B37 7110 7201 9917 9B0D 99E3 55DB 5D58 FADE 4A52





-----BEGIN PGP SIGNATURE-----
Version: PGP Desktop 9.8.0 (Build 2158)
Charset: us-ascii

wj8DBQFKyX+cSPXX9MTuPioRAjnrAKCGchbvHUNjio7rMiqPAoi0Kf7OawCghNsH
m8Jdj1okTfDg3Iq8uvkBOvw=
=xhok
-----END PGP SIGNATURE-----

[aix_tech]

Hello team,

This is the information you requeated found inside the two files listed below:

config.h

/* #undef HAVE_SYS_MTIO_H */


hostopts.h
/*-----------------------------------------------------------------*/
/* GNU 'C' (e.g. Linux) options...                                 */
/*-----------------------------------------------------------------*/
#define OPTION_SCSI_TAPE                /* SCSI tape support       */


I am really lost and will appreciate a recommendation!

Thank you

Victor



--- In hercules-390@..., "Fish" <fish@...> wrote:

>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> [...]
> > No, but I can tell you that when I went looking in
> > the gcc executable on my Linux system for something
> > to use for conditional compilation, I went with:
> >
> > defined(__gnu_linux__)
>
> Done! Thanks.
>
>
> - --
> "Fish" (David B. Trout) - fish@...
> Fight Spam! Join CAUCE! <http://www.cauce.org/>
> 7 reasons why HTML email is a bad thing
> http://www.georgedillon.com/web/html_email_is_evil.shtml
> PGP key fingerprints:
> DH/DSS: 9F9B BAB0 BA7F C458 1A89 FE26 48F5 D7F4 C4EE 3E2A
> RSA: 6B37 7110 7201 9917 9B0D 99E3 55DB 5D58 FADE 4A52
>
>
>
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: PGP Desktop 9.8.0 (Build 2158)
> Charset: us-ascii
>
> wj8DBQFKyX+cSPXX9MTuPioRAjnrAKCGchbvHUNjio7rMiqPAoi0Kf7OawCghNsH
> m8Jdj1okTfDg3Iq8uvkBOvw=
> =xhok
> -----END PGP SIGNATURE-----
>

[aix_tech]

Hi Paul,

To answer your question, I did not do anything to any of the config files. I am using all the defaults.

BTW, SCSI Tape support is there. AIX supports all known SCSI devices and, we actually have some SCSI tapes attached to this environment.

Perhaps I should not be taken the defaults.

One more thing, the hostopts.h does not contain a section for AIX, so I am assuming that the GNU and the LINUX sections might apply to us.

Your help will continue to be appreciated.

Thank you

Victor



--- In hercules-390@..., "kerravon86" <kerravon86@...> wrote:

>
> --- In hercules-390@..., "aix_tech" <aix_tech@> wrote:
> >
> > I am attempting to compile Hercules under AIX 5.3 and came
> > across the following errors, during the "make" phase.
> >  
> > hstructs.h:1133: error: field 'mtget' has incomplete type
> > hstructs.h:1147: error: field 'stape_getstat_mtget' has incomplete type make: 1254-004 The error code from the last command is 1.
> >
> > Can someone give comment on this?
>
> Did you switch on OPTION_SCSI_TAPE?  Try switching
> it off. My guess is that it's not supported on
> your system.
>
> BFN.  Paul.
>

[kerravon86]

Hi Victor. It's not your fault, it's Fish's. :-)

However, he's already fixed the problem, and it's
presumably in the snapshots already.

But regardless, he's also given you a workaround:

... you'll have to manually #undef OPTION_SCSI_TAPE in
"hostopts.h".

You won't be able to use SCSI, even though you
have SCSI, because it seems Hercules doesn't
support the SCSI method you use. Note - I don't
know anything about SCSI myself, so Fish is the
one who would have to answer what's going wrong.

I just know how to butcher C code. :-)

BFN.  Paul.





--- In hercules-390@..., "aix_tech" <aix_tech@...> wrote:

>
> Hi Paul,
>
> To answer your question, I did not do anything to any of the config files. I am using all the defaults.
>
> BTW, SCSI Tape support is there. AIX supports all known SCSI devices and, we actually have some SCSI tapes attached to this environment.
>
> Perhaps I should not be taken the defaults.
>
> One more thing, the hostopts.h does not contain a section for AIX, so I am assuming that the GNU and the LINUX sections might apply to us.
>
> Your help will continue to be appreciated.
>
> Thank you
>
> Victor
>
>
>
> --- In hercules-390@..., "kerravon86" <kerravon86@> wrote:
> >
> > --- In hercules-390@..., "aix_tech" <aix_tech@> wrote:
> > >
> > > I am attempting to compile Hercules under AIX 5.3 and came
> > > across the following errors, during the "make" phase.
> > >  
> > > hstructs.h:1133: error: field 'mtget' has incomplete type
> > > hstructs.h:1147: error: field 'stape_getstat_mtget' has incomplete type make: 1254-004 The error code from the last command is 1.
> > >
> > > Can someone give comment on this?
> >
> > Did you switch on OPTION_SCSI_TAPE?  Try switching
> > it off. My guess is that it's not supported on
> > your system.
> >
> > BFN.  Paul.
> >
>

[fish-8]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

kerravon86 wrote:

[...]
> You won't be able to use SCSI, even though you
> have SCSI, because it seems Hercules doesn't
> support the SCSI method you use.

Correct. For whatever reason the sys/mtio.h header is missing on your
system:

   /* #undef HAVE_SYS_MTIO_H */

Without that header Hercules is unable to provide support for SCSI.


> Note - I don't know anything about SCSI myself,
> so Fish is the one who would have to answer what's
>going wrong.

He's missing a required header for SCSI support.


- --
"Fish" (David B. Trout) - fish@...
Fight Spam! Join CAUCE! <http://www.cauce.org/>
7 reasons why HTML email is a bad thing
http://www.georgedillon.com/web/html_email_is_evil.shtml
PGP key fingerprints:
DH/DSS: 9F9B BAB0 BA7F C458 1A89 FE26 48F5 D7F4 C4EE 3E2A
RSA: 6B37 7110 7201 9917 9B0D 99E3 55DB 5D58 FADE 4A52





-----BEGIN PGP SIGNATURE-----
Version: PGP Desktop 9.8.0 (Build 2158)
Charset: us-ascii

wj8DBQFKzDVISPXX9MTuPioRAusIAJ9don7Q+oSvPZtGXHzOZSNH2fyO6ACfeuwJ
msbHTVE/j+sqJQR3zwTKncs=
=vd67
-----END PGP SIGNATURE-----

[Mike Schwab]

Could the LINUX section be copied to an AIX section and be modified as
reports come in?
Perhaps they have a different name for their SCSI header file?

On Wed, Oct 7, 2009 at 1:29 AM, Fish <fish@...> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> kerravon86 wrote:
>
> [...]
>> You won't be able to use SCSI, even though you
>> have SCSI, because it seems Hercules doesn't
>> support the SCSI method you use.
>
> Correct. For whatever reason the sys/mtio.h header is missing on your
> system:
>
>   /* #undef HAVE_SYS_MTIO_H */
>
> Without that header Hercules is unable to provide support for SCSI.
>
>
>> Note - I don't know anything about SCSI myself,
>> so Fish is the one who would have to answer what's
>>going wrong.
>
> He's missing a required header for SCSI support.
>
>
> - --
> "Fish" (David B. Trout) - fish@...

--
Mike A Schwab, Springfield IL USA
Where do Forest Rangers go to get away from it all?

[aix_tech]

Fish, Mike, Paul (Alpha order)

Gentlemen,

Thank you for your comments and suggestions.

I "#undef OPTION_SCSI_TAPE" and tried the "make" phase again.
This time I got the errors listed below.

Is there a place where I can get the latest copy of: hostopts.h ?

I am about to through the towels with the AIX compilation and just try to use Hercules under LINUX and/or Windows. I would hate it if I have to stick to Windows since it is not an OS I enjoy!

I also have several AIX servers at home and a large number at work. That's why I wanted to compile it under AIX and just copy the binary around.

My ultimate goal is to play with the EVAL version of z/VM, which I already have. But the path to that goal is full of stones and thorns!

Thank you

Victor


--- In hercules-390@..., Mike Schwab <Mike.A.Schwab@...> wrote:

>
> Could the LINUX section be copied to an AIX section and be modified as
> reports come in?
> Perhaps they have a different name for their SCSI header file?
>
> On Wed, Oct 7, 2009 at 1:29 AM, Fish <fish@...> wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > kerravon86 wrote:
> >
> > [...]
> >> You won't be able to use SCSI, even though you
> >> have SCSI, because it seems Hercules doesn't
> >> support the SCSI method you use.
> >
> > Correct. For whatever reason the sys/mtio.h header is missing on your
> > system:
> >
> >   /* #undef HAVE_SYS_MTIO_H */
> >
> > Without that header Hercules is unable to provide support for SCSI.
> >
> >
> >> Note - I don't know anything about SCSI myself,
> >> so Fish is the one who would have to answer what's
> >>going wrong.
> >
> > He's missing a required header for SCSI support.
> >
> >
> > - --
> > "Fish" (David B. Trout) - fish@...
>
> --
> Mike A Schwab, Springfield IL USA
> Where do Forest Rangers go to get away from it all?
>

[kerravon86]

--- In hercules-390@..., "aix_tech" <aix_tech@...> wrote:
>
> Fish, Mike, Paul (Alpha order)

Alphabetical order, I can accept.

Latitude order, I can't.

> I "#undef OPTION_SCSI_TAPE" and tried the "make" phase again.
> This time I got the errors listed below.

There was nothing listed below. So it presumably
compiled cleanly? Or did you send an attachment
in a group that doesn't allow attachments?

> Is there a place where I can get the latest copy of: hostopts.h ?

The daily snapshots?

> I am about to through the towels with the AIX
> compilation and just try to use Hercules under
> LINUX and/or Windows.

If it were me, I'd wait until you don't get any
useful responses before throwing in the towel.

> I would hate it if I have to stick to Windows
> since it is not an OS I enjoy!

Linux certainly gets more testing.

> I also have several AIX servers at home and a
> large number at work. That's why I wanted to
> compile it under AIX and just copy the binary
> around.

Sounds reasonable to me.

> My ultimate goal is to play with the EVAL version
> of z/VM, which I already have. But the path to that
> goal is full of stones and thorns!

Could be worse. You could be trying out this
new, apparently-untested environment a couple
of hours prior to giving a conference on it.

BFN.  Paul.

[aix_tech]

No, no. It did not compile correctly.
I thought I posted the errors. They are attached below.

Also, I do have the following fileset, as the output to the lslpp command shows:

lslpp -w /usr/include/sys/mtio.h
  File                                        Fileset               Type
  ----------------------------------------------------------------------------
  /usr/include/sys/mtio.h                     bos.adt.include       File

THE FOLLOWING ARE THE ERRORS FROM THE LAST COMPILATION ATTEMPT:

In file included from hstdinc.h:253,
                 from version.c:22:
hostopts.h:138:23: warning: extra tokens at end of #undef directive
In file included from hercules.h:109,
                 from version.c:27:
hmacros.h:206:4: warning: #warning Large File Support missing
        if /bin/sh ./libtool --tag=CC --mode=compile gcc -DHAVE_CONFIG_H  -I. -I. -I.  -I. -I./decNumber    -W -Wall -O3 -MT hscutl.lo -MD -MP -MF ".deps/hscutl.Tpo" -c -o hscutl.lo hscutl.c;  then mv -f ".deps/hscutl.Tpo" ".deps/hscutl.Plo"; else rm -f ".deps/hscutl.Tpo"; exit 1; fi
 gcc -DHAVE_CONFIG_H -I. -I. -I. -I. -I./decNumber -W -Wall -O3 -MT hscutl.lo -MD -MP -MF .deps/hscutl.Tpo -c hscutl.c  -DPIC -o .libs/hscutl.o
In file included from hstdinc.h:253,
                 from hscutl.c:42:
hostopts.h:138:23: warning: extra tokens at end of #undef directive
In file included from hercules.h:109,
                 from hscutl.c:47:
hmacros.h:206:4: warning: #warning Large File Support missing
hscutl.c: In function 'socket_keepalive':
hscutl.c:727: error: 'SOL_TCP' undeclared (first use in this function)
hscutl.c:727: error: (Each undeclared identifier is reported only once
hscutl.c:727: error: for each function it appears in.)
hscutl.c: In function 'hgets':
hscutl.c:784: warning: comparison is always false due to limited range of data type
make: 1254-004 The error code from the last command is 1.


Stop.
make: 1254-004 The error code from the last command is 1.


Stop.
make: 1254-004 The error code from the last command is 2.


Stop.

--- In hercules-390@..., "kerravon86" <kerravon86@...> wrote:

>
> --- In hercules-390@..., "aix_tech" <aix_tech@> wrote:
> >
> > Fish, Mike, Paul (Alpha order)
>
> Alphabetical order, I can accept.
>
> Latitude order, I can't.
>
> > I "#undef OPTION_SCSI_TAPE" and tried the "make" phase again.
> > This time I got the errors listed below.
>
> There was nothing listed below. So it presumably
> compiled cleanly? Or did you send an attachment
> in a group that doesn't allow attachments?
>
> > Is there a place where I can get the latest copy of: hostopts.h ?
>
> The daily snapshots?
>
> > I am about to through the towels with the AIX
> > compilation and just try to use Hercules under
> > LINUX and/or Windows.
>
> If it were me, I'd wait until you don't get any
> useful responses before throwing in the towel.
>
> > I would hate it if I have to stick to Windows
> > since it is not an OS I enjoy!
>
> Linux certainly gets more testing.
>
> > I also have several AIX servers at home and a
> > large number at work. That's why I wanted to
> > compile it under AIX and just copy the binary
> > around.
>
> Sounds reasonable to me.
>
> > My ultimate goal is to play with the EVAL version
> > of z/VM, which I already have. But the path to that
> > goal is full of stones and thorns!
>
> Could be worse. You could be trying out this
> new, apparently-untested environment a couple
> of hours prior to giving a conference on it.
>
> BFN.  Paul.
>

[Alexey Bozrikov]

a> Fish, Mike, Paul (Alpha order)

a> Gentlemen,

a> Thank you for your comments and suggestions.

a> I "#undef OPTION_SCSI_TAPE" and tried the "make" phase again.
a> This time I got the errors listed below.

a> Is there a place where I can get the latest copy of: hostopts.h ?

a> I am about to through the towels with the AIX compilation and just
a> try to use Hercules under LINUX and/or Windows. I would hate it if
a> I have to stick to Windows since it is not an OS I enjoy!

a> I also have several AIX servers at home and a large number at
a> work. That's why I wanted to compile it under AIX and just copy the binary around.

a> My ultimate goal is to play with the EVAL version of z/VM, which I
a> already have. But the path to that goal is full of stones and thorns!

a> Thank you

a> Victor

It seems that hercules can be compiled on AIX 5.3. I used IBM
VisualAge C/C++ version 7.0 (pretty old one). There are few things to
be considered, of course. First one is configure flags. I used as
follows:

 ./configure CFLAGS='-I/usr/local/include' CC='xlC -q64' LDFLAGS='-q64' AR='ar -X64' NM='nm -B -X64'

X64 and q64 flags should not be needed on 32-bit machine/version, I
presume. On 64-bit machine (POWER5 in my case) -m64 seems to be needed, as
it does not compile __asm__ directives in machdep.h

Then, I made change to hostopts.h, just add AIX as a separate host. I added
after linux section. I presume Z_OK is not needed (was my fault,
forgot zconf.h). SOL_TCP should be defined, as it is not in AIX
include files.

 /* AIX -- bozy */
#elif defined(_AIX)

#define DLL_IMPORT extern
#define DLL_EXPORT
#undef OPTION_SCSI_TAPE
#define  HOW_TO_IMPLEMENT_SH_COMMAND     USE_FORK_API_FOR_SH_COMMAND
#define OPTION_FBA_BLKDEVICE            /* FBA block device support  */
#define _LFS64_LARGEFILE 1
#define _ext_ppc
#define SOL_TCP IPPROTO_TCP
#ifndef Z_OK
#  define Z_OK 0
#endif
#define DEFAULT_HERCPRIO    0
#define DEFAULT_TOD_PRIO  -20
#define DEFAULT_CPU_PRIO   15
#define DEFAULT_DEV_PRIO    8
#define MAX_DEVICE_THREADS          255   /* (0 == unlimited)          */


Then just typed 'make', seems to work OK (lots of warnings of course, but same
with GCC). After building, hercules seem to start fine, VM/370 IPLs
fine and works. Did not try z/VM 5.3 evaluation, though. No networking
support (TUN/TAP not available on AIX).

Hope it helps.

Alex



a> --- In hercules-390@..., Mike Schwab <Mike.A.Schwab@...> wrote:

>>
>> Could the LINUX section be copied to an AIX section and be modified as
>> reports come in?
>> Perhaps they have a different name for their SCSI header file?
>>
>> On Wed, Oct 7, 2009 at 1:29 AM, Fish <fish@...> wrote:
>> > -----BEGIN PGP SIGNED MESSAGE-----
>> > Hash: SHA1
>> >
>> > kerravon86 wrote:
>> >
>> > [...]
>> >> You won't be able to use SCSI, even though you
>> >> have SCSI, because it seems Hercules doesn't
>> >> support the SCSI method you use.
>> >
>> > Correct. For whatever reason the sys/mtio.h header is missing on your
>> > system:
>> >
>> >   /* #undef HAVE_SYS_MTIO_H */
>> >
>> > Without that header Hercules is unable to provide support for SCSI.
>> >
>> >
>> >> Note - I don't know anything about SCSI myself,
>> >> so Fish is the one who would have to answer what's
>> >>going wrong.
>> >
>> > He's missing a required header for SCSI support.
>> >
>> >
>> > - --
>> > "Fish" (David B. Trout) - fish@...
>>
>> --
>> Mike A Schwab, Springfield IL USA
>> Where do Forest Rangers go to get away from it all?
>>

[aix_tech]

As suggested, I created the AIX section.
I am still getting:

hstructs.h:1133: error: field 'mtget' has incomplete type
hstructs.h:1147: error: field 'stape_getstat_mtget' has incomplete type
make: 1254-004 The error code from the last command is 1.


Stop.
make: 1254-004 The error code from the last command is 1.


Stop.
make: 1254-004 The error code from the last command is 2.


Stop.


--- In hercules-390@..., Alexey <bozy@...> wrote:

>
> a> Fish, Mike, Paul (Alpha order)
>
> a> Gentlemen,
>
> a> Thank you for your comments and suggestions.
>
> a> I "#undef OPTION_SCSI_TAPE" and tried the "make" phase again.
> a> This time I got the errors listed below.
>
> a> Is there a place where I can get the latest copy of: hostopts.h ?
>
> a> I am about to through the towels with the AIX compilation and just
> a> try to use Hercules under LINUX and/or Windows. I would hate it if
> a> I have to stick to Windows since it is not an OS I enjoy!
>
> a> I also have several AIX servers at home and a large number at
> a> work. That's why I wanted to compile it under AIX and just copy the binary around.
>
> a> My ultimate goal is to play with the EVAL version of z/VM, which I
> a> already have. But the path to that goal is full of stones and thorns!
>
> a> Thank you
>
> a> Victor
>
> It seems that hercules can be compiled on AIX 5.3. I used IBM
> VisualAge C/C++ version 7.0 (pretty old one). There are few things to
> be considered, of course. First one is configure flags. I used as
> follows:
>
>  ./configure CFLAGS='-I/usr/local/include' CC='xlC -q64' LDFLAGS='-q64' AR='ar -X64' NM='nm -B -X64'
>
> X64 and q64 flags should not be needed on 32-bit machine/version, I
> presume. On 64-bit machine (POWER5 in my case) -m64 seems to be needed, as
> it does not compile __asm__ directives in machdep.h
>
> Then, I made change to hostopts.h, just add AIX as a separate host. I added
> after linux section. I presume Z_OK is not needed (was my fault,
> forgot zconf.h). SOL_TCP should be defined, as it is not in AIX
> include files.
>
>  /* AIX -- bozy */
> #elif defined(_AIX)
>
> #define DLL_IMPORT extern
> #define DLL_EXPORT
> #undef OPTION_SCSI_TAPE
> #define  HOW_TO_IMPLEMENT_SH_COMMAND     USE_FORK_API_FOR_SH_COMMAND
> #define OPTION_FBA_BLKDEVICE            /* FBA block device support  */
> #define _LFS64_LARGEFILE 1
> #define _ext_ppc
> #define SOL_TCP IPPROTO_TCP
> #ifndef Z_OK
> #  define Z_OK 0
> #endif
> #define DEFAULT_HERCPRIO    0
> #define DEFAULT_TOD_PRIO  -20
> #define DEFAULT_CPU_PRIO   15
> #define DEFAULT_DEV_PRIO    8
> #define MAX_DEVICE_THREADS          255   /* (0 == unlimited)          */
>
>
> Then just typed 'make', seems to work OK (lots of warnings of course, but same
> with GCC). After building, hercules seem to start fine, VM/370 IPLs
> fine and works. Did not try z/VM 5.3 evaluation, though. No networking
> support (TUN/TAP not available on AIX).
>
> Hope it helps.
>
> Alex
>
>
>
> a> --- In hercules-390@..., Mike Schwab <Mike.A.Schwab@> wrote:
> >>
> >> Could the LINUX section be copied to an AIX section and be modified as
> >> reports come in?
> >> Perhaps they have a different name for their SCSI header file?
> >>
> >> On Wed, Oct 7, 2009 at 1:29 AM, Fish <fish@> wrote:
> >> > -----BEGIN PGP SIGNED MESSAGE-----
> >> > Hash: SHA1
> >> >
> >> > kerravon86 wrote:
> >> >
> >> > [...]
> >> >> You won't be able to use SCSI, even though you
> >> >> have SCSI, because it seems Hercules doesn't
> >> >> support the SCSI method you use.
> >> >
> >> > Correct. For whatever reason the sys/mtio.h header is missing on your
> >> > system:
> >> >
> >> >   /* #undef HAVE_SYS_MTIO_H */
> >> >
> >> > Without that header Hercules is unable to provide support for SCSI.
> >> >
> >> >
> >> >> Note - I don't know anything about SCSI myself,
> >> >> so Fish is the one who would have to answer what's
> >> >>going wrong.
> >> >
> >> > He's missing a required header for SCSI support.
> >> >
> >> >
> >> > - --
> >> > "Fish" (David B. Trout) - fish@
> >>
> >> --
> >> Mike A Schwab, Springfield IL USA
> >> Where do Forest Rangers go to get away from it all?
> >>
>

[kerravon86]

--- In hercules-390@..., "aix_tech" <aix_tech@...> wrote:
>
> Also, I do have the following fileset, as the output
rto the lslpp command shows:
>
> lslpp -w /usr/include/sys/mtio.h

In that case, maybe SCSI support can be enabled after
all. So the root cause of that problem then is
possibly configure not detecting your header file.

Are you able to do a "head" of it to make sure it
is readable?

> hostopts.h:138:23: warning: extra tokens at end of #undef directive

I think you have gone

#undef OPTION_SCSI_TAPE */

instead of

/* #undef OPTION_SCSI_TAPE */

when you commented out that SCSI thing. Or put
something similar to that anyway at the end of
the line.

> In file included from hercules.h:109,
>                  from version.c:27:
> hmacros.h:206:4: warning: #warning Large File Support missing

Looks like this build is going to restrict you to
2 GB disks.

> hscutl.c: In function 'socket_keepalive':
> hscutl.c:727: error: 'SOL_TCP' undeclared (first use in this function)

Looks like a programming error. Expecting a
SOL_TCP to be available, just because the
TCP_KEEPIDLE is available.

I don't know how to solve that properly. You
probably have a #define TCP_KEEPIDLE somewhere.
If you comment that out (same as SCSI tape), it
might work around the problem.

But if no-one else gives you a better answer than
that, yes, now is probably the time to throw in
the towel on the untested AIX environment. You
haven't even got to any possible runtime errors
yet. :-)

BFN.  Paul.


  #if defined(TCP_KEEPALIVE)
    optval = idle_time;
    rc = setsockopt(sfd, IPPROTO_TCP, TCP_KEEPALIVE, &optval, sizeof(optval));
    if (rc) logmsg("HHCUT002I TCP_KEEPALIVE rc=%d %s\n", rc, strerror(errno));
  #elif defined(TCP_KEEPIDLE)
    optval = idle_time;
    rc = setsockopt(sfd, SOL_TCP, TCP_KEEPIDLE, &optval, sizeof(optval));
    if (rc) logmsg("HHCUT003I TCP_KEEPIDLE rc=%d %s\n", rc, strerror(errno));
  #else

[kerravon86]

--- In hercules-390@..., "aix_tech" <aix_tech@...> wrote:
>
> As suggested, I created the AIX section.

I thought that suggestion was for the developers.

> I am still getting:
>
> hstructs.h:1133: error: field 'mtget' has incomplete type

You're still trying to use the SCSI. You have to
choose either:

1. undefine SCSI.
2. find the #undef HAVE_SYS_MTIO_H and instead make
it #define HAVE_SYS_MTIO_H, since configure doesn't
seem to be picking it up correctly.

Also, you now (sorry, missed it before) have an answer
from someone who got it working on AIX, so you should
probably persevere.

> >  ./configure CFLAGS='-I/usr/local/include'

I think this one is the main thing you need. If
you have this, then when you run configure, it will
hopefully put in that #define I mentioned above
properly.

And it may well work around that other problem too,
because it'll probably hit the first condition now.

BFN.  Paul.

[Alexey Bozrikov]

k> --- In hercules-390@..., "aix_tech" <aix_tech@...> wrote:
>>
>> As suggested, I created the AIX section.

k> I thought that suggestion was for the developers.


Not only for developers, although it is a good idea. Using this section,
Victor could possibly compile Hercules successfully. Don't know how far he can go with
GCC on AIX (I do not have it, so did not try), but it worked with VA C/C++.

>> I am still getting:
>>
>> hstructs.h:1133: error: field 'mtget' has incomplete type

k> You're still trying to use the SCSI. You have to
k> choose either:

k> 1. undefine SCSI.
k> 2. find the #undef HAVE_SYS_MTIO_H and instead make
k> it #define HAVE_SYS_MTIO_H, since configure doesn't
k> seem to be picking it up correctly.

Yes, this is what I was suggesting, too in my 'AIX' section. There is
mtio.h on AIX, but it can be way different from linux version.


k> Also, you now (sorry, missed it before) have an answer
k> from someone who got it working on AIX, so you should
k> probably persevere.

>> >  ./configure CFLAGS='-I/usr/local/include'

This was specifically for zlib/bzlib/zconf header files to enable DASD
compression. On Victor's system, since he is using GCC, this is most
probably included by default.

k> I think this one is the main thing you need. If
k> you have this, then when you run configure, it will
k> hopefully put in that #define I mentioned above
k> properly.

k> And it may well work around that other problem too,
k> because it'll probably hit the first condition now.

k> BFN.  Paul.




k> ------------------------------------

k> Community email addresses:
k>   Post message: hercules-390@...
k>   Subscribe:    hercules-390-subscribe@...
k>   Unsubscribe:  hercules-390-unsubscribe@...
k>   List owner:   hercules-390-owner@...

k> Files and archives at:
k>   http://groups.yahoo.com/group/hercules-390

k> Get the latest version of Hercules from:
k>   http://www.hercules-390.org

k> Yahoo! Groups Links






** Anything that begins well, ends badly.

[Alexey Bozrikov]

d> If I am not mistaken you are using gnu C, judging by the output
d> from your compiler. Have you tried XlC/visualAge ?

Would be reasonable, unless Victor is willing to go extra mile and get
it done with gcc. IBM has its compilers available with no-charge 30-day
trial license - more than enough to compile few projects.

To my experience with GCC on AIX 3.2.5, 4.3 and 5.1 - things could be tricky
with libraries and modified headers. It could be that GCC is having
trouble with mtio.h data types for this very reason (or, it could be
that GCC defines _LINUX_SOURCE_COMPAT macro, then struct mtget is not
defined in mtio.h).
VA C/C++ (even version 7) got enough GCC compatibility options, including __asm__
syntax processing to avoid use of GCC.

I just tried defining OPTION_SCSI_TAPE and seem to be getting same
error as Victor when compiling - no struct mtget defined. AIX might require few
more changes here and there to make SCSI tape support work. I do not
have any SCSI tapes on our AIX machines, thus cannot really test.

Just to note, that large file support can be enabled by defining
_LFS64_LARGEFILE.


Alexey




d> --- In hercules-390@..., "aix_tech" <aix_tech@...> wrote:

>>
>>
>> As suggested, I created the AIX section.
>> I am still getting:
>>
>> hstructs.h:1133: error: field 'mtget' has incomplete type
>> hstructs.h:1147: error: field 'stape_getstat_mtget' has incomplete type
>> make: 1254-004 The error code from the last command is 1.
>>
>> > It seems that hercules can be compiled on AIX 5.3. I used IBM
>> > VisualAge C/C++ version 7.0 (pretty old one). There are few things to

** Rainy days and automatic weapons always get me down.

[aix_tech]

Hi Paul,

I tried the suggestions listed here about the undefing of SCSI tape option and the MTIO. The make executed for a few more nano-seconds. I thought we got it, but at the end I still got the errors listed below.

Again, thank you for your continue support.

Victor

hmacros.h:206:4: warning: #warning Large File Support missing
In file included from hercules.h:150,
                 from hsys.c:26:
hstructs.h:1133: error: field 'mtget' has incomplete type
hstructs.h:1147: error: field 'stape_getstat_mtget' has incomplete type
make: 1254-004 The error code from the last command is 1.


Stop.
make: 1254-004 The error code from the last command is 1.


Stop.
make: 1254-004 The error code from the last command is 2.


Stop.

--- In hercules-390@..., "kerravon86" <kerravon86@...> wrote:

>
> --- In hercules-390@..., "aix_tech" <aix_tech@> wrote:
> >
> > As suggested, I created the AIX section.
>
> I thought that suggestion was for the developers.
>
> > I am still getting:
> >
> > hstructs.h:1133: error: field 'mtget' has incomplete type
>
> You're still trying to use the SCSI. You have to
> choose either:
>
> 1. undefine SCSI.
> 2. find the #undef HAVE_SYS_MTIO_H and instead make
> it #define HAVE_SYS_MTIO_H, since configure doesn't
> seem to be picking it up correctly.
>
> Also, you now (sorry, missed it before) have an answer
> from someone who got it working on AIX, so you should
> probably persevere.
>
> > >  ./configure CFLAGS='-I/usr/local/include'
>
> I think this one is the main thing you need. If
> you have this, then when you run configure, it will
> hopefully put in that #define I mentioned above
> properly.
>
> And it may well work around that other problem too,
> because it'll probably hit the first condition now.
>
> BFN.  Paul.
>